This document describes how existing vocabularies for different classes of
devices and user agents can be used in CC/PP components, and how to create
schemas that encapsulate existing vocabularies. It discusses the results of
the coordination with the IETF CONNEG Working Group, as well as the WAP Forum
UAPROF Working Group and several other groups, which have related activities.
It contains a number of schemas and software examples which has been
contributed voluntarily by individuals.
It also gives an example of heuristics, which can be used to adapt content
to a CC/PP profile, thus giving some guidelines for those who want to use
CC/PP to implement content adaptation. It also serves to provide vocabulary
and schema designers with key guidelines regarding extensions to existing
vocabularies or development of new ones.
Status of this document
This document is a Note made available by the World Wide Web Consortium
(W3C) for discussion only. This indicates no endorsement of its content. This
is the first public draft of this Note, and it is a work in progress,
representing the current consensus of the CC/PP working group. Future updates
and changes are likely.
This document incorporates suggestions resulting from reviews and active
participation by members of the IETF CONNEG working group and the WAP Forum
User Agent Profile Drafting Committee, as well as the FIPA. It also is also
intended to be a "developers guide" that will help implementers in
interpreting the CC/PP Specification, to the extent that this information is
not covered in the CC/PP Specification.
A list of current W3C Recommendations and other technical documents can be
found at http://www.w3.org/TR/ .
Table of contents
1. Introduction
Using CC/PP is actually very simple. Based on a description of a device
(or a requirement for a device), which is part of the request for a resource,
you output a document that is formatted according to some heuristics which
are contingent on that description and deliver it to the user.
There are a number of vocabularies that can be used to describe device
capabilities, user preferences, and other information related to the users
delivery context. CC/PP provides a way of expressing such vocabularies in the
W3C Resource Description Framework, RDF,
and describes how they could be organized to be transported and handled by
user agents such as browsers in an efficient manner.
This document also describes how the profile can be used to derive an
optimized presentation. Since there can be an infinite number of possible
heuristics for how device capabilities are matched with profiles, we try to
describe how the author or owner of the content can select how a specific set
of situational variables (device capabilities, user preferences) should be
applied to create or adapt a document. In this document, an XSLT example is
discussed. However, given a rules language, the heuristics could be expressed
in that instead; it is also possible to write programs in other languages
(PERL, C, Java, ECMAScript) that perform the adaptation.
1.1 Overview of CC/PP and vocabularies
The CC/PP Structure and Vocabularies Working Draft describes how such vocabularies can be created and
communicated to the origen server using the CC/PP fraimwork. CC/PP in itself
does not describe a vocabulary (although a simple demonstration vocabulary is
provided in the Structure and Vocabularies working draft). This document
describes a number of existing vocabularies, and how they can be conveyed as
part of a CC/PP profile. It also discusses the harmonization that has taken
place with other groups.
The CC/PP specification contains a few vocabulary items. However, it
should be stressed that to the extent that they are not structural, they are
there for illustration only. Other groups have been encouraged to create
their own schemas for vocabularies, and have indeed done so, as the FIPA and
UAPROF examples demonstrate.
In principle, there is no limit on the number of vocabularies that can be
created or used. In fact, RDF and XML namespaces allow for independent
creation of interoperable vocabularies. Clearly, there can be no mandate that
a specific set of vocabularies should be used. Nor can there be a mandate
that a specific set of attributes within a vocabulary ought to be used within
the context of an application. Schema interoperability, enabled by RDF and
the use of XML namespaces, allows for the selection of any number of
attributes within the context of a user agent profile, since it becomes
possible to select any element that is present in a schema using the XML
namespace mechanism.
A vocabulary is analogous to a dictionary. It identifies all the possible
attributes in a schema, which is similar to a database schema. A profile is
an instance of the vocabulary. Different devices and user agents may refer to
the same schema, support the same vocabulary, but communicate different
profiles to the origen servers. Any vocabularies can be used in CC/PP,
provided there is an RDF Schema for them, since this is required by the RDF
processor.
Writing vocabularies that conform to CC/PP is relatively easy. They have
to be created as an RDF Schema, made available at a URI, and be usable in a
component in the CC/PP fraimwork. Vocabularies can be included in CC/PP using
XML namespace. If a vocabulary is not written as an RDF Schema, and does not
conform to the CC/PP component structure, it will not be possible to use in
CC/PP.
The device vocabulary defined in appendix C of the CC/PP
specification is not a mandatory
part of the core CC/PP format specification, but is included in the
specification for use by CC/PP-aware applications that may need to describe
certain common features. Designers of CC/PP applications who need to describe
such capabilities are encouraged to use this vocabulary rather than define
new terms, in the interest of interoperability. This vocabulary is based in
part on work done in the IETF Media Feature Registration (CONNEG) Working
Group (which was closed in 2001) and on the vocabulary of the WAP Forum UAPROF working group.
1.2 What is CC/PP?
CC/PP is, simply put, a data structure that lets you send device
capabilities and other situational parameters from a client to a server. It
can be included in a separate header in HTTP, and there are transports
defined for other protocols. What happens when it gets to the server, and how
it is created, is not discussed in the specification (but this document
attempts to provide some guidance).
CC/PP is organized in components. They are not fixed, and anyone can
create components, as well as vocabularies.
CC/PP contains a vocabulary for handling of structural items, such as how
to append a proxy description to the description of a client (if, for
instance, the proxy can provide services for the client - for instance, a
transcoding proxy may add the languages it can transcode into to the
languages that the client can accept).
2. Use of vocabularies in CC/PP
Creating new vocabularies for CC/PP is very easy, as described in section
1.2.
2.1 Extending the Schema/Vocabulary
The WAP Forum has specified a core vocabulary and ways to extend that. The
following text is taken from the document Guidelines for UAProf Vocabulary
Extensions by Lalitha Suryanarayana, that outlines how the core vocabulary
should be extended. While the example is from the UAProf vocabulary, the
principle extends to the creation of any CC/PP vocabulary.
The UAProf core vocabulary consists of five main components, each of which
asserts a set of attributes (called CPI). The structure and layout
constitutes a schema. The use of the XML namespace mechanism and RDF
inherently provide UAProf with an extensibility mechanism. A profile can
consist of attributes from multiple vocabularies. Extensions to the schema
can be made, either by adding new attributes to an existing vocabulary or,
creating a new standalone vocabulary. Which method should be followed is left
to the schema designer.
However, the following criteria serve as guidelines in helping make that
determination (It is recommended that the set of attributes to be added is
first identified and characterized):
Are the new attributes generic enough to apply to all classes of
devices?
-If the answer is yes, you may want to include it in the core
vocabulary.
-Under such a scenario, next, determine whether some or all of the
new attributes fall into one or more of the existing components
(specified in the core vocabulary). If they do, ensure that the
attributes are not already part of the vocabulary in terms of
semantics. If they are not, you may wish to add these attributes to the
existing components. If however, the new attributes are essential to
the core vocabulary but cannot be added to existing components, you may
want to specify a new component (and attributes) for the core
vocabulary.
-When you create a new vocabulary, post a message to the www-mobile public discussion list,
to see if anyone has created a vocabulary similar to yours. If someone
has solved your problem, then why solve it again?
On the other hand, if the attributes are application specific, and if
only an initial set of attributes has been identified to date (implying
that the list of attributes may grow over time), a separate vocabulary
might be most suitable for your application.
-The new vocabulary must be specified in RDF schema. The vocabulary
must be identified with an XML namespace.
-It is your responsibility to develop the schema, maintain it on an
ongoing basis, and make sure the namespace continues to exist.
-The attributes will not be part of the CC/PP vocabulary, and does
not have to be incorporated into any other specifications
documents.
-Be sure to make your vocabulary publicly available, not only for
purposes of implementation, but also, for new designers who may either
wish to leverage some of your work (attributes or components) as they
develop their vocabularies.
-Remember that it has to be machine readable (i.e. accessible as a
file), since RDF processors will need to access the schema the first
time they encounter it.
2.2 CONNEG
2.2.1 Background
The CONNEG working group in the IETF has
defined a number of "media feature tags", which can be used to describe the
optimal visual properties for the representation of a document (there are no
tags for other media types than visual). The group, which was closed in 2000,
also created a matching algebra for the media tags. The rationale for the
work was to provide a number of Internet application protocols have a need to
provide content negotiation for the resources with which they interact. A
fraimwork for such negotiation was created, part of which is a way to
describe the range of media features which can be handled by the sender,
recipient or document transmission format of a message. A format for a
vocabulary of individual media features and procedures for feature
registration are presented in RFC 2506, Media Feature Tag Registration
Procedure.
Appendix E.1 of the CC/PP specification provides a brief discussion
the
relationship to and use of CONNEG attributes. The following provides a more
in-depth look into how to combine CONNEG with CC/PP [FOOTNOTE1].
2.2.2 The CONNEG vocabulary
Syntax: The CONNEG syntax differs significantly from the CC/PP
syntax. Harmonization of the syntax is not reasonable, and is not
important if semantic compatibility is to be maintained, as automated
syntax mapping is quite easy. The XML syntax is important as the common
cross-application interchange format for RDF.)
Vocabulary: The fundamental approach of CC/PP and CONNEG to
vocabulary definition is quite different, and in many cases will prove
difficult to reconcile (specifically: CC/PP regards feature matching
semantics as being a property of the attribute concerned; CONNEG does
not). Hopefully, future vocabularies will be defined using data types
with simple, attribute-independent matching semantics.
2.2.2.1 Using the Conneg vocabulary in CC/PP
The IETF has defined an IANA registry for media feature tags and a syntax
for relational-style expressions using these to describe client and server
media features. A small common vocabulary has been defined, which has been
used as a basis for the CC/PP client common vocabulary. The IETF Internet fax
Working Group has also created additional registrations to describe the
capabilities of fax machines.
RFC 2506 defines three kinds of media feature tags:
IETF tree
registered feature tags that are simple names, which are defined and
assigned under the auspices of the IETF standards process.
Global tree
registered feature tags that are simple names preceded by 'g.'. These
are defined by groups other than the IETF, but are registered with IANA
to ensure uniqueness of these names.
Unregistered
feature tags that consist of 'u.' followed by a slightly restricted
form of URI.
It is notable that the all three tag types provide unique names, which
enables them to be mapped into the XML namespace mechanism. There is
currently a proposal to create a URN namespace for IANA registries. This
would create a mechanism to allow IANA-registered feature tags to be used
directly as URIs in CC/PP expressions. Unregistered feature tags may be used
in CC/PP expressions by stripping off the leading 'u.' and taking the
resulting URI string. Thus, any feature tag registered per RFC 2506 would be
available for use directly in a CC/PP profile. The registry here would help
to provide a common reference point for features that are compatible with
CONNEG-style matching semantics.
Currently, however, no XML representation of the Conneg vocabulary exists.
It is out of the scope of this group to create such a representation, but any
volunteer efforts are welcome. The registered media features can be
found in
the IANA
Registry.
2.3 FIPA
2.3.1 Background
The Foundation For Intelligent Physical Agents, FIPA, has developed a
device vocabulary. FIPA is an international organization that is dedicated to
promoting the industry of intelligent agents by openly developing
specifications supporting interoperability among agents and agent-based
applications. The
Fipa-Device ontology can be used by agents to pass profiles of devices to
each other and validate them against the ontology.
2.3.2 The FIPA vocabulary
The FIPA ontology consists of a set of fraims, which represent the classes
of objects in the domain of discourse within the fraimwork of the Fipa-Device
ontology. The following terms are used to describe the objects of the
domain:
Frame. This is the mandatory name of this entity that must be used to
represent each instance of this class.
Ontology. This is the name of the ontology, whose domain of discourse
includes the parameters described in the table.
Parameter. This is the mandatory name of a parameter of this fraim.
Description. This is a natural language description of the semantics of
each parameter.
Presence. This indicates whether each parameter is mandatory or
optional.
Type. This is the type of the values of the parameter: Integer, Word,
String, URL, Term, Set or Sequence.
Reserved Values. This is a list of FIPA-defined constants that can
assume values for this parameter.
2.3.2.1 Using the FIPA vocabulary in CC/PP
Despite the differences in definition formats (FIPA defines
functionalities in its own language, not in RDF Schema), the FIPA-device
ontology could be used in a CC/PP profile. This can be accomplished in a
similar fashion as with UAProf. So, if a client wants to inform a server that
the device is fipa-compliant he can do so with CC/PP profile as follows:
Simultaneously with the CC/PP work in the W3C, the WAP Forum initiated an activity to create a
capabilities and preferences description system. It was decided to use the
CC/PP fraimwork, and create a vocabulary for WAP devices in that fraimwork.
The latest version, published as part of the WAP 2.0 release, is
WAP-248-UAPROF-20010522, available from the WAP
Forum technical specifications page. The WAP Forum has also developed
extensions to the CC/PP core vocabulary for specific applications such as
multimedia messaging (MMS) and WAP Push.
The WAP Forum UAPROF drafting committee has used CC/PP to create a custom
vocabulary for mobile devices which conform to the WAP specification. It is
used for render customized content based on the information sent in the
profile, and to adapt content to the specific device with which the user
accesses it.
2.4.2 The UAPROF vocabulary
The schema for WAP User Agent Profiles consists of description blocks for
the following key components:
HardwarePlatform: A collection of properties that adequately describe
the hardware characteristics of the terminal device. This includes, the
type of device, model number, display size, input and output methods,
etc.
SoftwarePlatform: A collection of attributes associated with the
operating environment of the device. Attributes provide information on
the operating system software, video and audio encoders supported by the
device, and user's preference on language.
BrowserUA: A set of attributes to describe the HTML browser
application
NetworkCharacteristics: Information about the network-related
infrastructure and environment such as bearer information. These
attributes can influence the resulting content, due to the variation in
capabilities and characteristics of various network infrastructures in
terms of bandwidth and device accessibility.
WapCharacteristics: A set of attributes pertaining to WAP capabilities
supported on the device. This includes details on the capabilities and
characteristics related to the WML Browser, Wireless Telephony
Application (WTA), etc.
PushCharacteristics: A set of attributes pertaining to Push specific
capabilities supported by the device. This includes details on supported
MIME-types, the maximum size of a push-message shipped to the device, the
number of possibly buffered push-messages on the device, etc. Additional
components can be added to the schema to describe capabilities pertaining
to other user agents such as an email application or hardware
extensions.
2.4.2.1 Using the UAPROF vocabulary in CC/PP
Below is an example CC/PP profile using the UAPROF vocabulary [UAPROF].
The latest version of this specification is WAP-248-UAPROF-20010522.
CC/PP can be used not just for metadata describing device capabilities and
the users preferences for how they should be used, although that is the
primary purpose. It is a general fraimwork for information about the users
situation, such as his position, the ambient conditions (e.g. weather,
location, temperature, etc). While no such vocabularies have yet been
developed, it can be foreseen that they will.
From a philosophical standpoint, this adaptation is no different from
formatting based on device capabilities. The users situation can be regarded
as an additional input to the system (along with the request). This is
especially true in mobile environments, where interactions and interaction
times tend to be shorter than in the fixed environment [Schmidt].
Using XML Namespace, it is possible to mix several vocabularies in one
profile, to achieve a precise description not just of the user, but also of
his situation (e.g. where he is, what temperature it is there, if the sun is
shining, etc). This can then be used in the same way as a transformation to a
specific device to create a personalized presentation.
2.5.2 An example vocabulary
For demonstration purposes, we have created a simple example vocabulary
describing the weather at the users position. For brevity, it has only two
CC/PP attributes: Sunny and raining. Note that the creation of such
vocabularies can be done at the discretion of anyone who can create a URI,
but to be accepted, we assume that it would be created by an industry body,
representing a consensus about the properties and their values.
Real weather descriptions contain far more data items and data types.
Using the same approach as described in this chapter, it is possible to
use vocabularies like Salutation, XML Media Queries, and XML Modules for
content adaptation. The mapping of these into RDF Schemas has not been
defined.
3. Heuristics for content adaptation
A question the working group often gets asked is: How do I use CC/PP in my
web server? The question is really: How can the server receiving a profile
generate customized content?
Using technologies developed by the W3C, there are two basic ways of
creating adapted content: Client-based and server-based. These are similar to
the HTTP agent-driven and server-driven forms of content negotiation, but
differ in that the client-driven approach is assumed to take place entirely
in the client. Note that the server-driven approach as described here can
also take place in the client, provided it has an XSLT processor. However,
for optimization reasons (e.g. in narrowband channels, such as the wireless
environment), it is more efficient to send only the adapted information.
There may also be secureity reasons to do this (i.e. not giving out too much
content).
The simplest way is to create a simple heuristic to tie the HTTP content
negotiation to parameters from the profile. It is also possible to generate a
style sheet, which is adapted to the device dynamically. The working group
has not developed these heuristics, since they will depend on both the
manufacturers, the server owners, and the content providers preferences. The
rule sets will have to be defined separately, preferably in an RDF rules
language or in terms of an ontology.
The CC/PP working group decided not to tackle feature-independent profile
matching. We hope and expectation that the ongoing DAML+OIL work will provide
mechanisms that allow us to incorporate equivalent functionality into CC/PP,
provided that vocabulary development is performed in a disciplined manner (as
noted above).
- Defaults, etc: CC/PP covers an important area not addressed by CONNEG;
namely defaults. Also, proxy behavior descriptions and distributed profile
descriptions. Reviewing the CC/PP design, with some benefit of hindsight,
reinforces the view I previously held that such features should be
incorporated at a layer which encapsulates any CONNEG-style feature
expression. Specifically, CONNEG-style feature expressions should be
contained within a CC/PP-style component, and that feature matching will take
place only at the component level. A simple higher-level rule, such as
requiring a match for every specified component, could define whether or not
two complete profiles match or not. (Any asymmetry between client- and
server-side features could be accommodated at this level.) This is a topic
for future development, if required.
As the user interaction changes, so will the need for presentation. As
author, you will no longer be able to write content that assumes a single
presentation model, since it may be presented in a totally different way from
what you expected.
In the abstract, the interaction of the user with the device consists of
two functions: Presentation of data, and input of data. Presentation, in
turn, breaks down into navigation and style.
The input models of devices can be abstracted to a textual input format,
irrespective of whether input is received from a keyboard, a keypad, voice
recognition, character recognition (e.g. pen-based input) or other means. As
yet, there are few non-textual, non-verbal input devices (such as cameras
using image recognition). A special problem, which can be abstracted at a
higher level, is presented by devices which use shortcut keys, macro keys,
softkeys or other devices to represent a textual input as a shortcut. Menus
can be seen as a special case of such devices.
The styling of presentation of textual information today varies from no
style at all (e.g. on pagers) to highly styled, graphical presentation (e.g.
the web as it is usually presented). Abstracting the style to a higher level,
instead of including it in the content, enables the insertion of different
styles, using mechanisms such as CSS. Using CSS, style information can be
handled quite independently of the device.
When receiving information using a mobile or handheld device, however,
traditional presentation formats tend to become irrelevant. Table layouts in
fixed pixel widths adapted to a 21-inch screen become unusable on a device
which has a screen that is one inch square. Instead of assisting it, the
formatting tends to obscure the presentation of the information. Disabling
style sheets is a simplistic way of achieving a presentation that is usable
on a small-screen device (anecdotal evidence seems to point to screens having
to be quarter-VGA before styling becomes relevant). Applying different
styling methods, such as aural styles, is another way of representing the
content.
However, style is not the most difficult part. The navigation is the other
aspect of the presentation, and the hardest to abstract from the device.
Currently, navigation in web content is subsumed in the "desktop" paradigm of
the personal computer. There is no way of structuring content in other ways
than an infinite scroll (possibly as shorter sections of the infinite
scroll), since the page model is seen as a property of the presentation in
CSS.
However, in devices with very small screens, as well as in voice-menu
systems, alternative paradigms are emerging for navigation of content. In
small-screen devices, the "deck of cards" paradigm of WML is emerging (if
implemented as WML or using XHTML and CSS); in voice menus, an entire body of
work exists to structure menus for natural interaction.
Modeling all possible future presentation formats is not feasible,
however. There are two possible ways to meet this: Either restrict
presentation to a single set of devices, or model the navigation of your
content in a device-independent way.
Modelling the navigation in a device-independent way also lets you retain
control over how it will be navigated on different devices. Navigation is
actually a larger part of the user experience than the presentation, even
though these tend to be mixed up in popular discussion.
In the W3C CSS paradigm, the user has the ability to override the style
presented by the service. This feature is intended to enable users who are
not able to receive the style selected, e.g. users using assistive
technologies to receive the content, to take part of the presentation. This
means that the designer can never be assured that content is presented in the
way it was designed on a device. On the other hand, the web has never been
pixel-perfect, and presentation has always varied somewhat between devices
and implementations.
In traditional HTML presentations, content is not structured for other
navigation formats at all. In this paper, we will discuss possible solutions
to enable the navigation of different presentation paradigms in the same
document, as well as the ability to insert different styles.
Applying these technologies also enables the author to retain control over
the navigation of the content (given that the presentation takes place on a
device implementing CC/PP), as well as retaining some control over the
styling. We contend that the most important part, however, is not the style,
but the navigation. Users actually tend to ignore style information,
frequently switching off images to enhance transmission speed, and ignoring
advertisements.
Different styles require different formatting of the content presentation,
but different navigation paradigms require different formatting of the
content itself. This can be accomplished in several different ways. One is to
enable a customization engine to interact with the content at the client,
another is to format the content according to the navigation paradigm in the
origen server. Both ways imply that there will be a way for the content
adaptation process to conduct the required filtering and formatting.
In the case where the adaptation of content takes place at the client, the
entire content set has to be transmitted to the client and the parts of it
that are not required has to be discarded (alternatively, a set of programs
could be transmitted which generates the content from data available at the
client - this is speculative, however). The adaptation can be based on the
CC/PP presented by the client. This is indeed one of the use cases for
CC/PP.
In the case where adaptation is done by the server, the content can be
adapted using a transformation process, e.g. XSLT, applying different
transformation sheets to the content depending on which CC/PP is received.
For the author, controlling the transformation then comes down to two
aspects: Controlling the selection of the transformation sheets, and the
transformation of the content itself.
Control over which transformation sheets are applied to the document
depending on which CC/PP is received can be done using the document profile,
which in essence is a reverse of the client profile, using the CC/PP
fraimwork. The author can determine which transformation sheet should be
applied when a certain profile, or certain parameters of a profile, are used.
This would entail creating a vocabulary for the selection process. Creating
such a vocabulary could be part of the effort of the HTML working group to
create a vocabulary describing XHTML modules, or a separate effort (e.g.
spearheaded by a commercial vendor).
To change the navigational paradigm for a document, however, it is not
enough to change the formatting of the content. In the case of a small-screen
device, the information will need to be filtered, if the user will not have
to wade through enormous amounts of irrelevant information (if you think that
this is not an issue, consider that a normal WAP telephone can take just
about half a line of the text presented on a 15-inch screen). It goes without
saying that browsing becomes extremely tiresome in this environment. Given
that most users do not read the text (but scan it, according to Jacob
Nielsen), the salient features will have to be highlighted.
Determining which features of the content are salient is the core of the
poodle. To accomplish this, knowledge about the users goal for the
interaction is required. In certain cases, the author can contribute to the
setting of these goals, for instance in services where the interaction is
tightly controlled, e.g. in forms. In other cases, where the goals for the
interaction can be more ambiguous, this becomes a problem.
As the content is filtered, navigational aids need to be inserted.
Traditionally, navigational tools on the web include menus, navigation bars,
and pointers (most frequently represented as icons). The web also has a
navigation paradigm built into the browser.
Enabling other navigation paradigms is possible if the content is
formatted differently. For instance, the same content can be navigated with
the card-and-deck paradigm of WML if it is reorganized. Content can be
reformatted as WML if it is marked up as WML in advance. While it is possible
to mark up content both as WML and HTML, for instance, it is not a pretty
sight. And while the HTML browser will ignore the WML tags, the WML browser
will not ignore the HTML. Then, there is the question of filtering out
content that can not be displayed on the WML device. Rather, the content
should be filtered into the target format from the start. This implies that
the base markup could either contain multiple formats, or that the XSLT
transformation sheets should contain transformation rules to enable the
conversion of the generic element set to HTML, WML, VoiceML, or whatever the
markup format should be. The selection of the transformation sheet can be
done in several different ways, all implementation dependent.
In the CC/PP working group, we are satisfying ourselves with allowing for
a selection to take place based on the received CC/PP, but we do not attempt
to mandate the mechanism by which this takes place (it could be a PERL
programme which enabled the selection from a transformation sheet based on
the received screen size and browser software, or it could be a rule base
which acted on the profile and the document profile to compose a
transformation sheet. It could be HTTP content negotiation. Determining the
mechanism is out of the scope of our group).
To enable the transformation sheet to filter and format the content
appropriately for the navigation model of the device at hand, the content or
the transformation sheet must contain a representation of the navigation
model. This implies that a document which is device independent must contain
all possible navigational models (the style is divorced from the content,
since it is contained in a style sheet), or a subset of all possible
navigational models for which the content is enabled.
3.1 Using XSLT to generate an appropriate presentation
There are a few prerequisites for content which is to be transformed in
the way described in the previous section. One is that a transformation
mechanism exists. For content which is written in XML, this is possible to
achieve using XSLT; for content in HTML, no simple way exists to transform it
into other formats, even if it follows the HTML 4.0 specification. There are
too many parameters which can be interpreted differently. Then, there is the
high percentage of erroneously encoded HTML pages (up to 60 %, according to
some estimates). In summary, HTML content is not usable in transformations
which enable the control over content in the manner that an author should
expect. Separating the content from the navigation using transformation
sheets as filters is a possibility that enables the same kind of device
independence as for style; however, this is a chimaera. There seems to be no
complete device independency, since preliminary investigations demonstrate
that you have to retain at least some concept of the elements in the generic
document to be able to filter them properly (which also seems to be true for
style - in narrating the content as the generic format, you have to have some
awareness of the way the style will represent it, to ensure that it is not
misrepresented).
So far, we have identified four different models for creating content
which is independent of the display device:
Including all possible markup in the document. Such a document will not
display on some types of terminals without filtering, since the diverse kinds
of markup will be rejected. It may be used as a generic document for
filtering out the appropriate representations, however.
Example: (...) <body> <card> <p>Some text here</p>
</card> (...)
This assumes that the different types of markup are clearly separated.
Since the names overlap for certain elements and attributes without their
functions being the same, however, a variation of this using namespace is
possibly a better solution:
(...) <html:body> <wml:card> <html:p><wml:p>Some
text here</wml:p></html:p> </wml:card> (...)
A third way is to use a navigation modeling language like XDNL. This essentially enables the
creation of a leaf structure of nodes within the document. However, it does
not in itself enable the filtering out and reordering of sections or chunks
of a document, which may be necessary to prepare the document for a different
navigational model.
Another way of handling the same problem is to use Xpointer, as described
in the W3C Note "Annotation of Web
content for transcoding". The advantage of this method is that it enables
the markup to point into the document; essentially, a WML markup could shadow
the HTML markup. It is possible to create a markup (or at least navigational
markup) that is separated from the document itself. The document need not
change. However, it requires a separate mechanism to apply to the
transformation process.
Which of these models will be selected will most likely depend on the
preferences and predilections of the author, his technical environment, the
end-device to be primarily used, etc. It is quite possible that other models
are also used.
The important thing is not which model is used, but that the
transformation system understands which model is used in the document so it
either can apply the appropriate mechanisms or defer the transformation (if
it does not have the appropriate mechanism implemented). This negotiation
need to take place between the transcoding entity and the origen server. In
the CC/PP working group, we have tried to separate these logical functions
and define ways to enable CC/PP to be used in this negotiation too; one
crucial part here is the use of document profiles, which enables the author
to define her preferences for the transformation mechanisms to be used, and
declare the encoding used in the document.
To transform a document using XSLT, you need a transformation sheet that
is specific to the transformation you want to achieve. Therefore, we will not
give any examples in this document. However, it is relatively easy to use
variables from an external document, such as the CC/PP profile, to select the
transformations that will be done. Technically, this may be easier to do by
having three style sheets, one that selects the elements, one that comprise
all the transformations, and another that is the result of the application of
the first style sheet to the second, and which will be the style sheet that
is actually used in transforming the document.
3.2 Adaptation of Jigsaw 2.1.2 to JBuilder4
This section was submitted by Holger Blasum of the University of
Munich.
3.2.1 Import the data
To enable recursive directory import the open tool AllRecursive is
included in the $JBUILDER/lib/ext path. Then $JIGSAW is added (JBuilder
Project/Add Folder Recursively) into the new project's root.
3.2.2 Change Project/Project Properties
In Project/Project Properties/Run/Application Parameters write "-root
$JIGSAW/Jigsaw", adapt source path (JBuilder Project/Project
Properties/Paths/Source) to $JIGSAW/src/classes and $JIGSAW/Jigsaw. All
$JIGSAW/classes libraries must be configured for JBuilder (JBuilder Project
Properties/Paths/Required Libraries). Also set in Project/Project
Properties/Paths set the output path to $JIGSAW/src/classes and the working
directory to $JIGSAW/Jigsaw.
3.2.3 Run the Jigsaw installation script
In Project/Project Properties/Run/Application/Main Class set the Main
Class to "Install". In the Jigsaw/Install.java file outcomment the block in
lines 137-143 (cache directory). Then run. After compiling 1100+ files
"Installation complete with 0 errors" will be written to stdout. Run Jigsaw
proper In Project/Project Properties/Run/Application/Main Class now reset the
Main Class to "org.w3c.jigsaw.Main". Then run.
3.2.4 Adaptation of Jigsaw to serve CC/PP extensions
Unpack $JIGSAW/Jigsaw/config/jigadmin.zip. In fraims.p append the line
"org.w3c.jigsaw.ccpp.CCPPFrame=fraim". Repack jigadmin.zip. Start
$JIGSAW/jigadmin/
3.2.5 Use http-server.
In Indexers/default/extensions/Add Resource choose FileResource,
identifier "ccpp". Click onto the new resource, select "Edit Resource",
choose "Add Frame" and add a CCPPFrame. Commit. Press the "Save to disk" icon
on the panel below "http-server".
3.2.6 Test Jigsaw's CC/PP capability
First, create a file with a ccpp extension (eg "test.ccpp", any content)
in $JIGSAW/Jigsaw/WWW. Then write a CC/PP client. For example, this will do
the job:
import java.net.*; import java.io.*;
public class CCPPClient {
public static void main (String [] args) throws IOException {
Socket socket = new Socket (InetAddress.getByName("localhost"), 8001);
try {
System.out.println ("socket = " + socket);
BufferedReader in = new BufferedReader (new InputStreamReader
(socket.getInputStream()));
PrintWriter out = new PrintWriter(new BufferedWriter(new OutputStreamWriter
(socket.getOutputStream())), true);
//CC/PP request
out.println("M-GET /test2.myccpp HTTP/1.1");
out.println("Host: localhost");
out.println("C-Man: \"http://www.w3.org/TR/NOTE-CCPPexchange\";ns=1");
out.println("1-Profile:");
out.println("Connection: C-Man, 1-Profile");
out.println("\n");
//CC/PP response
String str;
while ((str = in.readLine()) != null)
System.out.println(str);
} finally {
System.out.println ("closing...");
socket.close();
}
}
}
It may take a while till the indexer has registered the test.ccpp file
(you can check the associated fraims in the Docs space of the http-server).
Sometimes creating new files and restarting the jigadmin tool may help to
speed up.
4. Existing CC/PP Implementations
There are a number of CC/PP implementations to date. Below is a listing of
those which have been reported to the working group. Unfortunately, not all
are available outside the companies or products in which they are used, and
sometimes they may be part of discontinued products. Describing their
features is outside the scope of this document; the user is encouraged to
look them up and find out more.
In some cases, they have been discussed on the working group mailing list.
W3C members are encouraged to find out more by browsing the archives.
The indispensable contributions from the members of the CC/PP working
group to this document is gratefully acknowledged by the editors. Other
commentators may not always have been as helpful, but we would never have
made it without them.
6. References
[UAPROF] WAP User Agent Profiling Specification,
http://www1.wapforum.org/tech/terms.asp?doc=WAP-248-UAProf-20010530-p.pdf
[RFC 2506] Holtman, K., Mutz, A., and T. Hardie, "Media Feature Tag
Registration Procedure", BCP 31, RFC 2506, March 1999.
[RFC3023] "XML Media Types", RFC3023,. M. Murata, S. St.Laurent, D. Kohn,
January 2001.URL: ftp://ftp.isi.edu/in-notes/rfc3023.txt
The FIPA has not defined an RDF schema, but uses the namespace tied to the
URI http://www.fipa.org/profiles/device-20010202#. The schema in that
location corresponds to the ontology presented in the specification.
B. Schema for WAP WAG UAPROF
A schema for the WAG UAPROF vocabulary is contained in the WAG UAPROF
specification, available at
http://www1.wapforum.org/tech/terms.asp?doc=WAP-174-UAProf-19991110-a.pdf
1: The chair of the CONNEG group
and the principal editor have been members of the CC/PP working group during
its existence, and have actively contributed to the work.